There’s something unsettling about talking to a machine, even more so when you consider many don’t have a face.
Humans have spent centuries evolving to read micro-expressions, to gauge sincerity from the slight tightening around someone’s eyes, to build trust through eye contact. So when we strip away the visual layer and reduce interaction to text on a screen or a disembodied voice, we’re asking humans to communicate in a way that feels fundamentally unnatural.
That’s the problem Trulience set out to solve — and after sitting down with Richard Bowdler, their Head of Growth, I walked away convinced we’re witnessing something more profound than just better graphics. We’re watching the interface between humans and AI evolve from something transactional into something that might actually feel like a conversation.
The Story That Started It All
Most technology companies trace their origin to a technical insight or market opportunity. Trulience’s founding story is more personal than that, and honestly, it’s the kind of story that makes you remember why we build technology in the first place.
Marek, one of the founders, had already had a successful exit with video streaming technology to Blackboard, a major EdTech platform. He could have retired comfortably. Instead, he kept thinking about his elderly grandmother sitting in a care home, far from family, with limited interaction beyond occasional visits and brief conversations with overworked staff. The image haunted him — this woman who had lived a full life, now languishing with so little mental and emotional stimulation.
“I wonder if there’s a technical solution to this as well as a human solution,” Marek thought. That seed of an idea, planted about five years ago — well before the AI explosion we’re experiencing now — became Trulience.
What strikes me about this origin is how it frames the entire mission: not “how do we make better avatars,” but “how do we give AI a face and a body to make it more human?” It’s a subtle but crucial difference. The technology serves the human connection, not the other way around.
The Heavy Lifting of Early Days
When Richard described those early days, I couldn’t help but laugh in recognition. We’ve both been in this space long enough to remember when building XR experiences using Avatars required those massive 360-degree capture cages; a rig structure that positions an array of cameras surrounding a subject from every angle.
“Heavy studio builds, heavy lifts, full 360 cages of surround cameras and intensive filming days,” Richard recalled. Just creating a fairly rudimentary avatar required enormous front-end investment. You’d spend days capturing someone from every possible angle, hoping you had enough coverage to render realistic movement from any perspective.
But the visual challenge was only half the battle. The bigger problem? Making these avatars actually conversational.
“We were trying to get the avatars to be meaningfully and reasonably conversational,” Richard explained. “Now with the explosion of LLMs and the voice AI layers that sit on top, we’re kind of spoiled for choice. You can have conversations with AI and it feels incredibly fluid. But way back then, it was tricky.”
The technology has evolved dramatically even in the time Richard’s been with the company. They moved from capture cages to mocap suits — you know, those black suits covered in tracking dots — then shifted from building in Unreal Engine to more efficient techniques. The pace of advancement has been staggering, particularly in the last six months. Richard noted that the appetite for this technology has exploded specifically because the product has finally reached a level of maturity that makes it viable for real applications.
The Architecture: Bringing Your Own Brain
Here’s where it gets technically interesting. Trulience made a deliberate architectural decision that I think is brilliant: they don’t try to own the entire stack. Instead, they’ve positioned themselves as the visual front end — the face and body — while letting customers “bring your own brain.”
“We sit on top of the shoulders of LLM and voice AI giants,” Richard explained. “You can plug in any language model, any voice AI tech, and we play around with different combinations to see what really plays nicely together.”
This modularity means Trulience can focus on what they do best — creating hyper-realistic human avatars or characters (think off-trademark Mickey Mouse or a fluffy dog) — while leveraging best-in-class AI models for the conversational layer. They experiment with different combinations: various language models, different hosting options (private or public), and multiple versions to find the optimal fit. On the voice side, they test different speech-to-text and text-to-speech solutions to minimize latency and maximize naturalness.
But here’s a fascinating quirk Richard shared: there’s such a thing as too fast. They’ve integrated voice-to-voice technology, and through testing, they discovered that responses below about 500 milliseconds feel weirdly instant to humans. It breaks the illusion because real human conversations have natural pauses. We need that beat to process, to feel like we’re actually being heard rather than talking to a prediction machine.
“We have got to the point where we are as fast as we need to be. Faster than we need to be,” Richard said. It’s a delightful problem to have — throttling your system to feel more human.
The 90% Problem: Lip Sync
If you’ve ever watched a poorly dubbed foreign film, you know how jarring mismatched lip movements can be. Your brain can’t help but focus on it, and it destroys your ability to engage with the content. With AI avatars, lip sync is even more critical — and exponentially more complex.
“Lip sync and mouth is like 90% of it,” I said during our conversation, and Richard immediately agreed. “The second that lip sync kind of falls out, it breaks the magic.”
The technical challenge is enormous. Different sounds create different mouth shapes — what’s called “visemes” in the industry. The letter “m” requires closed lips. “O” creates a rounded opening. “Th” requires tongue between teeth. Now multiply that across dozens of languages, each with unique phonemes and mouth shapes. Arabic, Hindi, Mandarin — each language introduces new challenges.
Trulience has an entire team dedicated exclusively to lip syncing and mouth shapes. It’s not glamorous work, but it’s absolutely essential. Every avatar I’ve seen from them maintains that illusion because they’ve invested in getting this foundational element right.
For multilingual support, Trulience takes a smart approach: they don’t reinvent the wheel. Language detection and translation happen at the LLM layer, often through third-party providers. Trulience focuses on ensuring their avatars can accurately render the mouth shapes for whatever language the conversation demands. It’s another example of their modular philosophy — integrate the best tools for each job rather than building everything in-house.
Real-World Impact: Beyond the Demo
Technology is only as valuable as the problems it solves. So where are these interactive avatars actually making a difference?
Richard shared a story that stuck with me — a healthcare project in India working with illiterate populations. In many rural areas, literacy remains a significant barrier to accessing services, including healthcare. Text-based interfaces are useless. Even phone trees with numbered options don’t work if you can’t read or count reliably.
“We worked on a project in India where the avatars enabled illiterate users to access healthcare services using natural language,” Richard explained. An avatar speaks to the person in their local language, asking questions and guiding them through the process conversationally. The person responds naturally, and the avatar understands and adapts. No reading required. No complex navigation. Just human conversation.
This isn’t a niche use case. We’re talking about democratizing access to essential services for populations that have been systematically excluded by technology that assumes literacy.
The use cases extend across industries. E-commerce companies are deploying avatars for customer onboarding, replacing traditional form-based processes with conversational experiences. Users can ask questions, get clarification, and complete complex workflows without ever seeing a form field. The avatar guides them through it naturally.
Coaching and personal development is another area seeing significant adoption. Imagine a fitness coach or business mentor who can interact with thousands of clients simultaneously, each receiving personalized attention. The expert records their knowledge and personality once, and the avatar becomes a scalable version of their expertise.
“The potential of avatars to democratize access to expert knowledge and free up time for coaches and creators” is massive, Richard noted. A leadership coach can work with C-suite executives who can afford their $10,000/month rate, while their avatar handles the $99/month tier, making that same expertise accessible to small business owners and entrepreneurs. The expert makes money while they sleep, and more people benefit from knowledge that would otherwise be limited by the expert’s available hours.
Customer service is perhaps the most obvious application, but also one of the most impactful. Instead of navigating phone trees or chat interfaces, customers interact with an avatar that can see their account, understand their history, and actually solve problems. The avatar has infinite patience, never has a bad day, and can handle the same question for the ten-thousandth time with the same enthusiasm as the first.
The Competitive Edge: Client-Side Rendering
Every technical founder loves talking about their differentiator, and when I asked Richard what sets Trulience apart, his answer was immediate: client-side rendering.
“Our key differentiator is client-side rendering of avatars, making the platform significantly cheaper and more scalable than competitors,” Richard explained. This is a big deal, and worth unpacking.
Most avatar platforms use server-side rendering, which means every frame of video is generated on the company’s servers and streamed to the user. This approach requires massive computational resources — think high-end GPUs churning away to render every blink, every lip movement, every subtle facial expression. The costs scale linearly with usage. More users mean more servers, more GPUs, more infrastructure.
Trulience flipped this model. They render avatars on the user’s device — your phone, your laptop, your browser. The avatar data and instructions are small enough to transmit efficiently, and modern devices are powerful enough to handle the rendering locally. The cost savings are dramatic. Instead of scaling server infrastructure with every new user, Trulience leverages the billions of devices already in users’ hands.
The scalability implications are enormous. A competitor might pay $X per minute of avatar interaction. Trulience pays a fraction of that because they’re not shouldering the computational burden. They can offer more competitive pricing, higher margins, or both.
This architectural decision also improves latency and reliability. Data doesn’t have to make round trips to distant servers. The rendering happens locally, which means faster response times and fewer points of failure. If your internet connection hiccups, the avatar continues functioning because it’s running on your device.
Staying Ahead in a Competitive Space
The avatar space is getting crowded. New companies pop up weekly, all promising the next generation of human-AI interaction. So how does a company like Trulience maintain its edge?
“We focus on customer success, reliable partners, and a solid product strategy,” Richard said. It sounds simple, almost too simple. But watching how they operate, it makes sense.
Customer success means obsessing over actual deployment and results rather than just closing deals. They work closely with early customers to understand what works, what breaks, and what features actually move the needle. This feedback loop informs product development and ensures they’re building for real needs rather than theoretical use cases.
Reliable partners matter because, as we’ve discussed, Trulience doesn’t try to own the entire stack. Their success depends on integrating seamlessly with best-in-class LLMs, voice AI providers, and other infrastructure. Choosing partners carefully — and maintaining those relationships — ensures the end-to-end experience remains solid even as individual components evolve.
The product strategy comes down to focus. “We need a period of narrowing and narrowing focus,” Richard admitted. “Classic thinking, crossing the chasm thesis is always go an inch wide and a mile deep. Pick our beachhead and nail it and dominate in one space. Then that earns you the right to go after others with more firepower, with more ease.”
This is mature thinking from a growth leader. The temptation in a hot market is to chase every opportunity, to be everything to everyone. But that’s how companies lose their way. Better to dominate one vertical — prove the value, build case studies, create repeatable processes — then expand from strength.
Flexible Integration: No-Code to Full Control
One question I always ask: how do developers actually integrate this technology? The answer matters because it determines who can build with your platform.
Trulience supports the full spectrum. On one end, they offer no-code tools — drag-and-drop interfaces that let non-technical teams build avatar experiences. Marketing can create a customer service avatar without filing an engineering ticket. Product managers can prototype conversational flows without writing a line of code.
On the other end, they provide comprehensive APIs and SDKs for developers who need deeper control. Want to customize the avatar’s appearance programmatically? There’s an API for that. Need to inject custom logic into the conversation flow? You can. Want to integrate with your existing user authentication and data systems? Fully supported.
“We support both no-code and developer-friendly integrations, from drag-and-drop tools to APIs and SDKs for deeper control,” Richard explained. This flexibility means Trulience can serve both enterprises with dedicated engineering teams and smaller companies that need to move fast with limited technical resources.
The philosophy mirrors their “bring your own brain” approach — give people the tools and get out of their way. Some customers will use everything out of the box. Others will customize extensively. Both are valid, and both are supported.
Ethics, Children, and Responsible AI
As a parent, Richard brings a unique perspective to the ethics of AI avatars. We talked about the responsibility that comes with building technology that can convincingly impersonate human interaction.
“As a parent and advocate, I stress the importance of ethical AI development and transparency in avatars,” Richard said. This isn’t just corporate responsibility theater. When you’re building technology that children will interact with, the stakes are high.
Transparency is crucial. Users should understand when they’re talking to an AI avatar rather than a human. This seems obvious, but in practice, it requires deliberate design choices. Visual indicators, clear labeling, and explicit disclosures all matter. The goal isn’t to trick people into thinking they’re talking to a human — it’s to create an interface that feels natural while remaining honest about what it is.
For children specifically, the considerations multiply. Kids are more impressionable, less able to distinguish reality from simulation, and more vulnerable to manipulation. Richard advocates for age-appropriate experiences, parental controls, and content restrictions that protect children while still allowing them to benefit from the technology.
There’s also the question of data privacy. Avatar conversations generate rich datasets about users — their concerns, their questions, their emotional states. This data is valuable but sensitive. Responsible platforms need robust privacy protections, clear data policies, and genuine user consent.
The industry is still figuring this out, but I appreciate that Trulience is thinking about these questions now rather than after a scandal forces the conversation.
The PayPal Moment
One of the most memorable moments in our conversation was when Richard compared current avatar adoption to the early days of online payments.
“Where kind of pre-PayPal, really, online payments existed in people’s minds to where online payments now sit in people’s minds, we are in a sort of pre-PayPal era when it comes to human-computer, human-avatar interactions right now,” he said.
Think about that analogy. In the late 1990s, typing your credit card number into a website felt dangerous, almost foolish. People were terrified of online fraud. They’d drive to a store rather than risk an online transaction. Banks and merchants hadn’t figured out security and fraud protection. The whole ecosystem was immature.
Then PayPal came along — not because they had perfect security (they didn’t), but because they made online payments feel safe enough, convenient enough, and ubiquitous enough that people were willing to try. Once they tried and it worked, the behavior normalized. Today, we don’t think twice about making online payments. We store our payment information everywhere. The risk hasn’t disappeared, but our comfort level has fundamentally shifted.
We’re in that pre-PayPal moment with AI avatars. For some people, the technology is so novel it’s fun — a curiosity to play with. For others, it’s scary or off-putting. “Loads of people out there have no idea that this kind of technology does exist,” Richard noted.
But as people gain experience with LLMs through ChatGPT and similar interfaces, they’re developing comfort with AI conversation. Adding a face and body is the logical next step. The interaction becomes more natural, more engaging, more human. And just like online payments, once people try it and it works, the behavior will normalize.
“People will inevitably become comfortable interacting with LLMs. And they are already, like with ChatGPT. But interacting with humanoids, avatars — that is a place that people will move to,” Richard predicted.
I think he’s right. We’re watching a fundamental shift in human-computer interaction, and we’re still early enough that developers and companies building in this space will have outsized impact on how it evolves.
The Unexpected Passion: Memory and Neurofeedback
I always ask guests what area of AI they’d explore if they weren’t in their current role. Richard’s answer took me completely by surprise.
“I have a background in human memory development. So in a former life, I went down a rabbit hole of going and doing memory competitions and these sorts of things,” he said. Memory competitions — the people who memorize decks of cards or recite pi to thousands of digits.
But it gets better. Richard also ran a project involving neurofeedback — non-invasive brain-computer interfaces that read emotional states and generate audio in real-time. The goal was to navigate through audio landscapes by controlling and navigating through different emotional states, with brain signals correlating to different audio outputs.
“Super cool, super fun to run,” Richard said with genuine enthusiasm. “I’d probably possibly be spending time doing something in that arena.”
This is someone who lives on the frontier of human-computer interaction in the most literal sense — using brain signals to control computers, exploring how memory works, and now giving AI a human face. It makes sense that these interests converge. They’re all about making technology more human, more intuitive, more natural.
“It’s my happy place right on the front. Just doing weird stuff right on the edge,” Richard laughed.
Looking Forward
As our conversation wound down, I found myself thinking about the broader implications of this technology. We’re not just making prettier chatbots. We’re fundamentally changing how humans and AI interact, and that change has cascading effects across education, healthcare, commerce, entertainment, and accessibility.
The technology Trulience is building isn’t the end state — Richard was clear about that. “I don’t think there’s that much more to do on the visual hyperrealism side of things on a foundational level. There’s the nuances, like lip syncing and so on. But then really it’s like, let’s scale. Let’s distribute and drive adoption.”
The product is ready. The question now is adoption and distribution. How do we move from early adopters and proof-of-concepts to mainstream deployment? How do we educate the market about the possibilities while building trust and addressing legitimate concerns?
For developers reading this, the opportunity is clear. The infrastructure is maturing, the APIs exist, and the use cases are proven. You can start building avatar experiences today, not years from now. The companies and developers who figure out compelling applications now will have significant first-mover advantages.
For everyone else, the message is simpler: this technology isn’t science fiction anymore. It’s real, it’s available, and it’s starting to show up in places you interact with every day. That avatar helping you navigate your insurance options? Powered by technology like Trulience’s. The personal coach that’s available 24/7? Same deal. The educational tutor that adapts to your learning style? You guessed it.
We’re giving AI a face, and in doing so, we’re making technology more human. That feels like progress worth pursuing.
Watch my discussion with Richard in its entirety:
You can explore Trulience’s technology yourself at trulience.com and interact with their pre-built avatars. For developers interested in building with Agora’s Conversational AI infrastructure, visit agora.io to see how our real-time communication platform powers these kinds of immersive experiences.
